src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_single/nb_kernel_ElecEwSw_VdwNone_GeomW4W4_avx_128_fma_single.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_128_fma_single kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_128_fma_single.h"
  48 #include "kernelutil_x86_avx_128_fma_single.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwNone_GeomW4W4_VF_avx_128_fma_single
  52  * Electrostatics interaction: Ewald
  53  * VdW interaction:            None
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecEwSw_VdwNone_GeomW4W4_VF_avx_128_fma_single
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB,jnrC,jnrD;
  75     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  76     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  81     real             scratch[4*DIM];
  82     __m128           fscal,rcutoff,rcutoff2,jidxall;
  83     int              vdwioffset1;
  84     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  85     int              vdwioffset2;
  86     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  87     int              vdwioffset3;
  88     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  89     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  90     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  91     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  92     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  93     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
  94     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  95     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  96     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  97     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
  98     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  99     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 100     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 101     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 102     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 103     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 104     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
 105     real             *charge;
 106     __m128i          ewitab;
 107     __m128           ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 108     __m128           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 109     real             *ewtab;
 110     __m128           rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 111     real             rswitch_scalar,d_scalar;
 112     __m128           dummy_mask,cutoff_mask;
 113     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 114     __m128           one     = _mm_set1_ps(1.0);
 115     __m128           two     = _mm_set1_ps(2.0);
 116     x                = xx[0];
 117     f                = ff[0];
 118
 119     nri              = nlist->nri;
 120     iinr             = nlist->iinr;
 121     jindex           = nlist->jindex;
 122     jjnr             = nlist->jjnr;
 123     shiftidx         = nlist->shift;
 124     gid              = nlist->gid;
 125     shiftvec         = fr->shift_vec[0];
 126     fshift           = fr->fshift[0];
 127     facel            = _mm_set1_ps(fr->epsfac);
 128     charge           = mdatoms->chargeA;
 129
 130     sh_ewald         = _mm_set1_ps(fr->ic->sh_ewald);
 131     beta             = _mm_set1_ps(fr->ic->ewaldcoeff_q);
 132     beta2            = _mm_mul_ps(beta,beta);
 133     beta3            = _mm_mul_ps(beta,beta2);
 134     ewtab            = fr->ic->tabq_coul_FDV0;
 135     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
 136     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
 137
 138     /* Setup water-specific parameters */
 139     inr              = nlist->iinr[0];
 140     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
 141     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
 142     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
 143
 144     jq1              = _mm_set1_ps(charge[inr+1]);
 145     jq2              = _mm_set1_ps(charge[inr+2]);
 146     jq3              = _mm_set1_ps(charge[inr+3]);
 147     qq11             = _mm_mul_ps(iq1,jq1);
 148     qq12             = _mm_mul_ps(iq1,jq2);
 149     qq13             = _mm_mul_ps(iq1,jq3);
 150     qq21             = _mm_mul_ps(iq2,jq1);
 151     qq22             = _mm_mul_ps(iq2,jq2);
 152     qq23             = _mm_mul_ps(iq2,jq3);
 153     qq31             = _mm_mul_ps(iq3,jq1);
 154     qq32             = _mm_mul_ps(iq3,jq2);
 155     qq33             = _mm_mul_ps(iq3,jq3);
 156
 157     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 158     rcutoff_scalar   = fr->rcoulomb;
 159     rcutoff          = _mm_set1_ps(rcutoff_scalar);
 160     rcutoff2         = _mm_mul_ps(rcutoff,rcutoff);
 161
 162     rswitch_scalar   = fr->rcoulomb_switch;
 163     rswitch          = _mm_set1_ps(rswitch_scalar);
 164     /* Setup switch parameters */
 165     d_scalar         = rcutoff_scalar-rswitch_scalar;
 166     d                = _mm_set1_ps(d_scalar);
 167     swV3             = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
 168     swV4             = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 169     swV5             = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 170     swF2             = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
 171     swF3             = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 172     swF4             = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 173
 174     /* Avoid stupid compiler warnings */
 175     jnrA = jnrB = jnrC = jnrD = 0;
 176     j_coord_offsetA = 0;
 177     j_coord_offsetB = 0;
 178     j_coord_offsetC = 0;
 179     j_coord_offsetD = 0;
 180
 181     outeriter        = 0;
 182     inneriter        = 0;
 183
 184     for(iidx=0;iidx<4*DIM;iidx++)
 185     {
 186         scratch[iidx] = 0.0;
 187     }
 188
 189     /* Start outer loop over neighborlists */
 190     for(iidx=0; iidx<nri; iidx++)
 191     {
 192         /* Load shift vector for this list */
 193         i_shift_offset   = DIM*shiftidx[iidx];
 194
 195         /* Load limits for loop over neighbors */
 196         j_index_start    = jindex[iidx];
 197         j_index_end      = jindex[iidx+1];
 198
 199         /* Get outer coordinate index */
 200         inr              = iinr[iidx];
 201         i_coord_offset   = DIM*inr;
 202
 203         /* Load i particle coords and add shift vector */
 204         gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 205                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 206
 207         fix1             = _mm_setzero_ps();
 208         fiy1             = _mm_setzero_ps();
 209         fiz1             = _mm_setzero_ps();
 210         fix2             = _mm_setzero_ps();
 211         fiy2             = _mm_setzero_ps();
 212         fiz2             = _mm_setzero_ps();
 213         fix3             = _mm_setzero_ps();
 214         fiy3             = _mm_setzero_ps();
 215         fiz3             = _mm_setzero_ps();
 216
 217         /* Reset potential sums */
 218         velecsum         = _mm_setzero_ps();
 219
 220         /* Start inner kernel loop */
 221         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 222         {
 223
 224             /* Get j neighbor index, and coordinate index */
 225             jnrA             = jjnr[jidx];
 226             jnrB             = jjnr[jidx+1];
 227             jnrC             = jjnr[jidx+2];
 228             jnrD             = jjnr[jidx+3];
 229             j_coord_offsetA  = DIM*jnrA;
 230             j_coord_offsetB  = DIM*jnrB;
 231             j_coord_offsetC  = DIM*jnrC;
 232             j_coord_offsetD  = DIM*jnrD;
 233
 234             /* load j atom coordinates */
 235             gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 236                                               x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 237                                               &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 238
 239             /* Calculate displacement vector */
 240             dx11             = _mm_sub_ps(ix1,jx1);
 241             dy11             = _mm_sub_ps(iy1,jy1);
 242             dz11             = _mm_sub_ps(iz1,jz1);
 243             dx12             = _mm_sub_ps(ix1,jx2);
 244             dy12             = _mm_sub_ps(iy1,jy2);
 245             dz12             = _mm_sub_ps(iz1,jz2);
 246             dx13             = _mm_sub_ps(ix1,jx3);
 247             dy13             = _mm_sub_ps(iy1,jy3);
 248             dz13             = _mm_sub_ps(iz1,jz3);
 249             dx21             = _mm_sub_ps(ix2,jx1);
 250             dy21             = _mm_sub_ps(iy2,jy1);
 251             dz21             = _mm_sub_ps(iz2,jz1);
 252             dx22             = _mm_sub_ps(ix2,jx2);
 253             dy22             = _mm_sub_ps(iy2,jy2);
 254             dz22             = _mm_sub_ps(iz2,jz2);
 255             dx23             = _mm_sub_ps(ix2,jx3);
 256             dy23             = _mm_sub_ps(iy2,jy3);
 257             dz23             = _mm_sub_ps(iz2,jz3);
 258             dx31             = _mm_sub_ps(ix3,jx1);
 259             dy31             = _mm_sub_ps(iy3,jy1);
 260             dz31             = _mm_sub_ps(iz3,jz1);
 261             dx32             = _mm_sub_ps(ix3,jx2);
 262             dy32             = _mm_sub_ps(iy3,jy2);
 263             dz32             = _mm_sub_ps(iz3,jz2);
 264             dx33             = _mm_sub_ps(ix3,jx3);
 265             dy33             = _mm_sub_ps(iy3,jy3);
 266             dz33             = _mm_sub_ps(iz3,jz3);
 267
 268             /* Calculate squared distance and things based on it */
 269             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 270             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 271             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 272             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 273             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 274             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 275             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 276             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 277             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 278
 279             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 280             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 281             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 282             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 283             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 284             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 285             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 286             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 287             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 288
 289             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 290             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 291             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
 292             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 293             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 294             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
 295             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
 296             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
 297             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
 298
 299             fjx1             = _mm_setzero_ps();
 300             fjy1             = _mm_setzero_ps();
 301             fjz1             = _mm_setzero_ps();
 302             fjx2             = _mm_setzero_ps();
 303             fjy2             = _mm_setzero_ps();
 304             fjz2             = _mm_setzero_ps();
 305             fjx3             = _mm_setzero_ps();
 306             fjy3             = _mm_setzero_ps();
 307             fjz3             = _mm_setzero_ps();
 308
 309             /**************************
 310              * CALCULATE INTERACTIONS *
 311              **************************/
 312
 313             if (gmx_mm_any_lt(rsq11,rcutoff2))
 314             {
 315
 316             r11              = _mm_mul_ps(rsq11,rinv11);
 317
 318             /* EWALD ELECTROSTATICS */
 319
 320             /* Analytical PME correction */
 321             zeta2            = _mm_mul_ps(beta2,rsq11);
 322             rinv3            = _mm_mul_ps(rinvsq11,rinv11);
 323             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 324             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 325             felec            = _mm_mul_ps(qq11,felec);
 326             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 327             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv11);
 328             velec            = _mm_mul_ps(qq11,velec);
 329
 330             d                = _mm_sub_ps(r11,rswitch);
 331             d                = _mm_max_ps(d,_mm_setzero_ps());
 332             d2               = _mm_mul_ps(d,d);
 333             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 334
 335             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 336
 337             /* Evaluate switch function */
 338             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 339             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv11,_mm_mul_ps(velec,dsw)) );
 340             velec            = _mm_mul_ps(velec,sw);
 341             cutoff_mask      = _mm_cmplt_ps(rsq11,rcutoff2);
 342
 343             /* Update potential sum for this i atom from the interaction with this j atom. */
 344             velec            = _mm_and_ps(velec,cutoff_mask);
 345             velecsum         = _mm_add_ps(velecsum,velec);
 346
 347             fscal            = felec;
 348
 349             fscal            = _mm_and_ps(fscal,cutoff_mask);
 350
 351              /* Update vectorial force */
 352             fix1             = _mm_macc_ps(dx11,fscal,fix1);
 353             fiy1             = _mm_macc_ps(dy11,fscal,fiy1);
 354             fiz1             = _mm_macc_ps(dz11,fscal,fiz1);
 355
 356             fjx1             = _mm_macc_ps(dx11,fscal,fjx1);
 357             fjy1             = _mm_macc_ps(dy11,fscal,fjy1);
 358             fjz1             = _mm_macc_ps(dz11,fscal,fjz1);
 359
 360             }
 361
 362             /**************************
 363              * CALCULATE INTERACTIONS *
 364              **************************/
 365
 366             if (gmx_mm_any_lt(rsq12,rcutoff2))
 367             {
 368
 369             r12              = _mm_mul_ps(rsq12,rinv12);
 370
 371             /* EWALD ELECTROSTATICS */
 372
 373             /* Analytical PME correction */
 374             zeta2            = _mm_mul_ps(beta2,rsq12);
 375             rinv3            = _mm_mul_ps(rinvsq12,rinv12);
 376             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 377             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 378             felec            = _mm_mul_ps(qq12,felec);
 379             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 380             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv12);
 381             velec            = _mm_mul_ps(qq12,velec);
 382
 383             d                = _mm_sub_ps(r12,rswitch);
 384             d                = _mm_max_ps(d,_mm_setzero_ps());
 385             d2               = _mm_mul_ps(d,d);
 386             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 387
 388             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 389
 390             /* Evaluate switch function */
 391             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 392             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv12,_mm_mul_ps(velec,dsw)) );
 393             velec            = _mm_mul_ps(velec,sw);
 394             cutoff_mask      = _mm_cmplt_ps(rsq12,rcutoff2);
 395
 396             /* Update potential sum for this i atom from the interaction with this j atom. */
 397             velec            = _mm_and_ps(velec,cutoff_mask);
 398             velecsum         = _mm_add_ps(velecsum,velec);
 399
 400             fscal            = felec;
 401
 402             fscal            = _mm_and_ps(fscal,cutoff_mask);
 403
 404              /* Update vectorial force */
 405             fix1             = _mm_macc_ps(dx12,fscal,fix1);
 406             fiy1             = _mm_macc_ps(dy12,fscal,fiy1);
 407             fiz1             = _mm_macc_ps(dz12,fscal,fiz1);
 408
 409             fjx2             = _mm_macc_ps(dx12,fscal,fjx2);
 410             fjy2             = _mm_macc_ps(dy12,fscal,fjy2);
 411             fjz2             = _mm_macc_ps(dz12,fscal,fjz2);
 412
 413             }
 414
 415             /**************************
 416              * CALCULATE INTERACTIONS *
 417              **************************/
 418
 419             if (gmx_mm_any_lt(rsq13,rcutoff2))
 420             {
 421
 422             r13              = _mm_mul_ps(rsq13,rinv13);
 423
 424             /* EWALD ELECTROSTATICS */
 425
 426             /* Analytical PME correction */
 427             zeta2            = _mm_mul_ps(beta2,rsq13);
 428             rinv3            = _mm_mul_ps(rinvsq13,rinv13);
 429             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 430             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 431             felec            = _mm_mul_ps(qq13,felec);
 432             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 433             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv13);
 434             velec            = _mm_mul_ps(qq13,velec);
 435
 436             d                = _mm_sub_ps(r13,rswitch);
 437             d                = _mm_max_ps(d,_mm_setzero_ps());
 438             d2               = _mm_mul_ps(d,d);
 439             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 440
 441             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 442
 443             /* Evaluate switch function */
 444             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 445             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv13,_mm_mul_ps(velec,dsw)) );
 446             velec            = _mm_mul_ps(velec,sw);
 447             cutoff_mask      = _mm_cmplt_ps(rsq13,rcutoff2);
 448
 449             /* Update potential sum for this i atom from the interaction with this j atom. */
 450             velec            = _mm_and_ps(velec,cutoff_mask);
 451             velecsum         = _mm_add_ps(velecsum,velec);
 452
 453             fscal            = felec;
 454
 455             fscal            = _mm_and_ps(fscal,cutoff_mask);
 456
 457              /* Update vectorial force */
 458             fix1             = _mm_macc_ps(dx13,fscal,fix1);
 459             fiy1             = _mm_macc_ps(dy13,fscal,fiy1);
 460             fiz1             = _mm_macc_ps(dz13,fscal,fiz1);
 461
 462             fjx3             = _mm_macc_ps(dx13,fscal,fjx3);
 463             fjy3             = _mm_macc_ps(dy13,fscal,fjy3);
 464             fjz3             = _mm_macc_ps(dz13,fscal,fjz3);
 465
 466             }
 467
 468             /**************************
 469              * CALCULATE INTERACTIONS *
 470              **************************/
 471
 472             if (gmx_mm_any_lt(rsq21,rcutoff2))
 473             {
 474
 475             r21              = _mm_mul_ps(rsq21,rinv21);
 476
 477             /* EWALD ELECTROSTATICS */
 478
 479             /* Analytical PME correction */
 480             zeta2            = _mm_mul_ps(beta2,rsq21);
 481             rinv3            = _mm_mul_ps(rinvsq21,rinv21);
 482             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 483             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 484             felec            = _mm_mul_ps(qq21,felec);
 485             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 486             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv21);
 487             velec            = _mm_mul_ps(qq21,velec);
 488
 489             d                = _mm_sub_ps(r21,rswitch);
 490             d                = _mm_max_ps(d,_mm_setzero_ps());
 491             d2               = _mm_mul_ps(d,d);
 492             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 493
 494             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 495
 496             /* Evaluate switch function */
 497             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 498             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv21,_mm_mul_ps(velec,dsw)) );
 499             velec            = _mm_mul_ps(velec,sw);
 500             cutoff_mask      = _mm_cmplt_ps(rsq21,rcutoff2);
 501
 502             /* Update potential sum for this i atom from the interaction with this j atom. */
 503             velec            = _mm_and_ps(velec,cutoff_mask);
 504             velecsum         = _mm_add_ps(velecsum,velec);
 505
 506             fscal            = felec;
 507
 508             fscal            = _mm_and_ps(fscal,cutoff_mask);
 509
 510              /* Update vectorial force */
 511             fix2             = _mm_macc_ps(dx21,fscal,fix2);
 512             fiy2             = _mm_macc_ps(dy21,fscal,fiy2);
 513             fiz2             = _mm_macc_ps(dz21,fscal,fiz2);
 514
 515             fjx1             = _mm_macc_ps(dx21,fscal,fjx1);
 516             fjy1             = _mm_macc_ps(dy21,fscal,fjy1);
 517             fjz1             = _mm_macc_ps(dz21,fscal,fjz1);
 518
 519             }
 520
 521             /**************************
 522              * CALCULATE INTERACTIONS *
 523              **************************/
 524
 525             if (gmx_mm_any_lt(rsq22,rcutoff2))
 526             {
 527
 528             r22              = _mm_mul_ps(rsq22,rinv22);
 529
 530             /* EWALD ELECTROSTATICS */
 531
 532             /* Analytical PME correction */
 533             zeta2            = _mm_mul_ps(beta2,rsq22);
 534             rinv3            = _mm_mul_ps(rinvsq22,rinv22);
 535             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 536             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 537             felec            = _mm_mul_ps(qq22,felec);
 538             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 539             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv22);
 540             velec            = _mm_mul_ps(qq22,velec);
 541
 542             d                = _mm_sub_ps(r22,rswitch);
 543             d                = _mm_max_ps(d,_mm_setzero_ps());
 544             d2               = _mm_mul_ps(d,d);
 545             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 546
 547             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 548
 549             /* Evaluate switch function */
 550             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 551             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv22,_mm_mul_ps(velec,dsw)) );
 552             velec            = _mm_mul_ps(velec,sw);
 553             cutoff_mask      = _mm_cmplt_ps(rsq22,rcutoff2);
 554
 555             /* Update potential sum for this i atom from the interaction with this j atom. */
 556             velec            = _mm_and_ps(velec,cutoff_mask);
 557             velecsum         = _mm_add_ps(velecsum,velec);
 558
 559             fscal            = felec;
 560
 561             fscal            = _mm_and_ps(fscal,cutoff_mask);
 562
 563              /* Update vectorial force */
 564             fix2             = _mm_macc_ps(dx22,fscal,fix2);
 565             fiy2             = _mm_macc_ps(dy22,fscal,fiy2);
 566             fiz2             = _mm_macc_ps(dz22,fscal,fiz2);
 567
 568             fjx2             = _mm_macc_ps(dx22,fscal,fjx2);
 569             fjy2             = _mm_macc_ps(dy22,fscal,fjy2);
 570             fjz2             = _mm_macc_ps(dz22,fscal,fjz2);
 571
 572             }
 573
 574             /**************************
 575              * CALCULATE INTERACTIONS *
 576              **************************/
 577
 578             if (gmx_mm_any_lt(rsq23,rcutoff2))
 579             {
 580
 581             r23              = _mm_mul_ps(rsq23,rinv23);
 582
 583             /* EWALD ELECTROSTATICS */
 584
 585             /* Analytical PME correction */
 586             zeta2            = _mm_mul_ps(beta2,rsq23);
 587             rinv3            = _mm_mul_ps(rinvsq23,rinv23);
 588             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 589             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 590             felec            = _mm_mul_ps(qq23,felec);
 591             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 592             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv23);
 593             velec            = _mm_mul_ps(qq23,velec);
 594
 595             d                = _mm_sub_ps(r23,rswitch);
 596             d                = _mm_max_ps(d,_mm_setzero_ps());
 597             d2               = _mm_mul_ps(d,d);
 598             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 599
 600             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 601
 602             /* Evaluate switch function */
 603             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 604             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv23,_mm_mul_ps(velec,dsw)) );
 605             velec            = _mm_mul_ps(velec,sw);
 606             cutoff_mask      = _mm_cmplt_ps(rsq23,rcutoff2);
 607
 608             /* Update potential sum for this i atom from the interaction with this j atom. */
 609             velec            = _mm_and_ps(velec,cutoff_mask);
 610             velecsum         = _mm_add_ps(velecsum,velec);
 611
 612             fscal            = felec;
 613
 614             fscal            = _mm_and_ps(fscal,cutoff_mask);
 615
 616              /* Update vectorial force */
 617             fix2             = _mm_macc_ps(dx23,fscal,fix2);
 618             fiy2             = _mm_macc_ps(dy23,fscal,fiy2);
 619             fiz2             = _mm_macc_ps(dz23,fscal,fiz2);
 620
 621             fjx3             = _mm_macc_ps(dx23,fscal,fjx3);
 622             fjy3             = _mm_macc_ps(dy23,fscal,fjy3);
 623             fjz3             = _mm_macc_ps(dz23,fscal,fjz3);
 624
 625             }
 626
 627             /**************************
 628              * CALCULATE INTERACTIONS *
 629              **************************/
 630
 631             if (gmx_mm_any_lt(rsq31,rcutoff2))
 632             {
 633
 634             r31              = _mm_mul_ps(rsq31,rinv31);
 635
 636             /* EWALD ELECTROSTATICS */
 637
 638             /* Analytical PME correction */
 639             zeta2            = _mm_mul_ps(beta2,rsq31);
 640             rinv3            = _mm_mul_ps(rinvsq31,rinv31);
 641             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 642             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 643             felec            = _mm_mul_ps(qq31,felec);
 644             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 645             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv31);
 646             velec            = _mm_mul_ps(qq31,velec);
 647
 648             d                = _mm_sub_ps(r31,rswitch);
 649             d                = _mm_max_ps(d,_mm_setzero_ps());
 650             d2               = _mm_mul_ps(d,d);
 651             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 652
 653             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 654
 655             /* Evaluate switch function */
 656             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 657             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv31,_mm_mul_ps(velec,dsw)) );
 658             velec            = _mm_mul_ps(velec,sw);
 659             cutoff_mask      = _mm_cmplt_ps(rsq31,rcutoff2);
 660
 661             /* Update potential sum for this i atom from the interaction with this j atom. */
 662             velec            = _mm_and_ps(velec,cutoff_mask);
 663             velecsum         = _mm_add_ps(velecsum,velec);
 664
 665             fscal            = felec;
 666
 667             fscal            = _mm_and_ps(fscal,cutoff_mask);
 668
 669              /* Update vectorial force */
 670             fix3             = _mm_macc_ps(dx31,fscal,fix3);
 671             fiy3             = _mm_macc_ps(dy31,fscal,fiy3);
 672             fiz3             = _mm_macc_ps(dz31,fscal,fiz3);
 673
 674             fjx1             = _mm_macc_ps(dx31,fscal,fjx1);
 675             fjy1             = _mm_macc_ps(dy31,fscal,fjy1);
 676             fjz1             = _mm_macc_ps(dz31,fscal,fjz1);
 677
 678             }
 679
 680             /**************************
 681              * CALCULATE INTERACTIONS *
 682              **************************/
 683
 684             if (gmx_mm_any_lt(rsq32,rcutoff2))
 685             {
 686
 687             r32              = _mm_mul_ps(rsq32,rinv32);
 688
 689             /* EWALD ELECTROSTATICS */
 690
 691             /* Analytical PME correction */
 692             zeta2            = _mm_mul_ps(beta2,rsq32);
 693             rinv3            = _mm_mul_ps(rinvsq32,rinv32);
 694             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 695             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 696             felec            = _mm_mul_ps(qq32,felec);
 697             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 698             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv32);
 699             velec            = _mm_mul_ps(qq32,velec);
 700
 701             d                = _mm_sub_ps(r32,rswitch);
 702             d                = _mm_max_ps(d,_mm_setzero_ps());
 703             d2               = _mm_mul_ps(d,d);
 704             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 705
 706             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 707
 708             /* Evaluate switch function */
 709             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 710             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv32,_mm_mul_ps(velec,dsw)) );
 711             velec            = _mm_mul_ps(velec,sw);
 712             cutoff_mask      = _mm_cmplt_ps(rsq32,rcutoff2);
 713
 714             /* Update potential sum for this i atom from the interaction with this j atom. */
 715             velec            = _mm_and_ps(velec,cutoff_mask);
 716             velecsum         = _mm_add_ps(velecsum,velec);
 717
 718             fscal            = felec;
 719
 720             fscal            = _mm_and_ps(fscal,cutoff_mask);
 721
 722              /* Update vectorial force */
 723             fix3             = _mm_macc_ps(dx32,fscal,fix3);
 724             fiy3             = _mm_macc_ps(dy32,fscal,fiy3);
 725             fiz3             = _mm_macc_ps(dz32,fscal,fiz3);
 726
 727             fjx2             = _mm_macc_ps(dx32,fscal,fjx2);
 728             fjy2             = _mm_macc_ps(dy32,fscal,fjy2);
 729             fjz2             = _mm_macc_ps(dz32,fscal,fjz2);
 730
 731             }
 732
 733             /**************************
 734              * CALCULATE INTERACTIONS *
 735              **************************/
 736
 737             if (gmx_mm_any_lt(rsq33,rcutoff2))
 738             {
 739
 740             r33              = _mm_mul_ps(rsq33,rinv33);
 741
 742             /* EWALD ELECTROSTATICS */
 743
 744             /* Analytical PME correction */
 745             zeta2            = _mm_mul_ps(beta2,rsq33);
 746             rinv3            = _mm_mul_ps(rinvsq33,rinv33);
 747             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 748             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 749             felec            = _mm_mul_ps(qq33,felec);
 750             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 751             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv33);
 752             velec            = _mm_mul_ps(qq33,velec);
 753
 754             d                = _mm_sub_ps(r33,rswitch);
 755             d                = _mm_max_ps(d,_mm_setzero_ps());
 756             d2               = _mm_mul_ps(d,d);
 757             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 758
 759             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 760
 761             /* Evaluate switch function */
 762             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 763             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv33,_mm_mul_ps(velec,dsw)) );
 764             velec            = _mm_mul_ps(velec,sw);
 765             cutoff_mask      = _mm_cmplt_ps(rsq33,rcutoff2);
 766
 767             /* Update potential sum for this i atom from the interaction with this j atom. */
 768             velec            = _mm_and_ps(velec,cutoff_mask);
 769             velecsum         = _mm_add_ps(velecsum,velec);
 770
 771             fscal            = felec;
 772
 773             fscal            = _mm_and_ps(fscal,cutoff_mask);
 774
 775              /* Update vectorial force */
 776             fix3             = _mm_macc_ps(dx33,fscal,fix3);
 777             fiy3             = _mm_macc_ps(dy33,fscal,fiy3);
 778             fiz3             = _mm_macc_ps(dz33,fscal,fiz3);
 779
 780             fjx3             = _mm_macc_ps(dx33,fscal,fjx3);
 781             fjy3             = _mm_macc_ps(dy33,fscal,fjy3);
 782             fjz3             = _mm_macc_ps(dz33,fscal,fjz3);
 783
 784             }
 785
 786             fjptrA             = f+j_coord_offsetA;
 787             fjptrB             = f+j_coord_offsetB;
 788             fjptrC             = f+j_coord_offsetC;
 789             fjptrD             = f+j_coord_offsetD;
 790
 791             gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 792                                                    fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 793
 794             /* Inner loop uses 477 flops */
 795         }
 796
 797         if(jidx<j_index_end)
 798         {
 799
 800             /* Get j neighbor index, and coordinate index */
 801             jnrlistA         = jjnr[jidx];
 802             jnrlistB         = jjnr[jidx+1];
 803             jnrlistC         = jjnr[jidx+2];
 804             jnrlistD         = jjnr[jidx+3];
 805             /* Sign of each element will be negative for non-real atoms.
 806              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 807              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 808              */
 809             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 810             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 811             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 812             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 813             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 814             j_coord_offsetA  = DIM*jnrA;
 815             j_coord_offsetB  = DIM*jnrB;
 816             j_coord_offsetC  = DIM*jnrC;
 817             j_coord_offsetD  = DIM*jnrD;
 818
 819             /* load j atom coordinates */
 820             gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 821                                               x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 822                                               &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 823
 824             /* Calculate displacement vector */
 825             dx11             = _mm_sub_ps(ix1,jx1);
 826             dy11             = _mm_sub_ps(iy1,jy1);
 827             dz11             = _mm_sub_ps(iz1,jz1);
 828             dx12             = _mm_sub_ps(ix1,jx2);
 829             dy12             = _mm_sub_ps(iy1,jy2);
 830             dz12             = _mm_sub_ps(iz1,jz2);
 831             dx13             = _mm_sub_ps(ix1,jx3);
 832             dy13             = _mm_sub_ps(iy1,jy3);
 833             dz13             = _mm_sub_ps(iz1,jz3);
 834             dx21             = _mm_sub_ps(ix2,jx1);
 835             dy21             = _mm_sub_ps(iy2,jy1);
 836             dz21             = _mm_sub_ps(iz2,jz1);
 837             dx22             = _mm_sub_ps(ix2,jx2);
 838             dy22             = _mm_sub_ps(iy2,jy2);
 839             dz22             = _mm_sub_ps(iz2,jz2);
 840             dx23             = _mm_sub_ps(ix2,jx3);
 841             dy23             = _mm_sub_ps(iy2,jy3);
 842             dz23             = _mm_sub_ps(iz2,jz3);
 843             dx31             = _mm_sub_ps(ix3,jx1);
 844             dy31             = _mm_sub_ps(iy3,jy1);
 845             dz31             = _mm_sub_ps(iz3,jz1);
 846             dx32             = _mm_sub_ps(ix3,jx2);
 847             dy32             = _mm_sub_ps(iy3,jy2);
 848             dz32             = _mm_sub_ps(iz3,jz2);
 849             dx33             = _mm_sub_ps(ix3,jx3);
 850             dy33             = _mm_sub_ps(iy3,jy3);
 851             dz33             = _mm_sub_ps(iz3,jz3);
 852
 853             /* Calculate squared distance and things based on it */
 854             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 855             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 856             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 857             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 858             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 859             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 860             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 861             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 862             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 863
 864             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 865             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 866             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 867             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 868             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 869             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 870             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 871             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 872             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 873
 874             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 875             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 876             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
 877             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 878             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 879             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
 880             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
 881             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
 882             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
 883
 884             fjx1             = _mm_setzero_ps();
 885             fjy1             = _mm_setzero_ps();
 886             fjz1             = _mm_setzero_ps();
 887             fjx2             = _mm_setzero_ps();
 888             fjy2             = _mm_setzero_ps();
 889             fjz2             = _mm_setzero_ps();
 890             fjx3             = _mm_setzero_ps();
 891             fjy3             = _mm_setzero_ps();
 892             fjz3             = _mm_setzero_ps();
 893
 894             /**************************
 895              * CALCULATE INTERACTIONS *
 896              **************************/
 897
 898             if (gmx_mm_any_lt(rsq11,rcutoff2))
 899             {
 900
 901             r11              = _mm_mul_ps(rsq11,rinv11);
 902             r11              = _mm_andnot_ps(dummy_mask,r11);
 903
 904             /* EWALD ELECTROSTATICS */
 905
 906             /* Analytical PME correction */
 907             zeta2            = _mm_mul_ps(beta2,rsq11);
 908             rinv3            = _mm_mul_ps(rinvsq11,rinv11);
 909             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 910             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 911             felec            = _mm_mul_ps(qq11,felec);
 912             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 913             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv11);
 914             velec            = _mm_mul_ps(qq11,velec);
 915
 916             d                = _mm_sub_ps(r11,rswitch);
 917             d                = _mm_max_ps(d,_mm_setzero_ps());
 918             d2               = _mm_mul_ps(d,d);
 919             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 920
 921             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 922
 923             /* Evaluate switch function */
 924             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 925             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv11,_mm_mul_ps(velec,dsw)) );
 926             velec            = _mm_mul_ps(velec,sw);
 927             cutoff_mask      = _mm_cmplt_ps(rsq11,rcutoff2);
 928
 929             /* Update potential sum for this i atom from the interaction with this j atom. */
 930             velec            = _mm_and_ps(velec,cutoff_mask);
 931             velec            = _mm_andnot_ps(dummy_mask,velec);
 932             velecsum         = _mm_add_ps(velecsum,velec);
 933
 934             fscal            = felec;
 935
 936             fscal            = _mm_and_ps(fscal,cutoff_mask);
 937
 938             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 939
 940              /* Update vectorial force */
 941             fix1             = _mm_macc_ps(dx11,fscal,fix1);
 942             fiy1             = _mm_macc_ps(dy11,fscal,fiy1);
 943             fiz1             = _mm_macc_ps(dz11,fscal,fiz1);
 944
 945             fjx1             = _mm_macc_ps(dx11,fscal,fjx1);
 946             fjy1             = _mm_macc_ps(dy11,fscal,fjy1);
 947             fjz1             = _mm_macc_ps(dz11,fscal,fjz1);
 948
 949             }
 950
 951             /**************************
 952              * CALCULATE INTERACTIONS *
 953              **************************/
 954
 955             if (gmx_mm_any_lt(rsq12,rcutoff2))
 956             {
 957
 958             r12              = _mm_mul_ps(rsq12,rinv12);
 959             r12              = _mm_andnot_ps(dummy_mask,r12);
 960
 961             /* EWALD ELECTROSTATICS */
 962
 963             /* Analytical PME correction */
 964             zeta2            = _mm_mul_ps(beta2,rsq12);
 965             rinv3            = _mm_mul_ps(rinvsq12,rinv12);
 966             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 967             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 968             felec            = _mm_mul_ps(qq12,felec);
 969             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 970             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv12);
 971             velec            = _mm_mul_ps(qq12,velec);
 972
 973             d                = _mm_sub_ps(r12,rswitch);
 974             d                = _mm_max_ps(d,_mm_setzero_ps());
 975             d2               = _mm_mul_ps(d,d);
 976             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 977
 978             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 979
 980             /* Evaluate switch function */
 981             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 982             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv12,_mm_mul_ps(velec,dsw)) );
 983             velec            = _mm_mul_ps(velec,sw);
 984             cutoff_mask      = _mm_cmplt_ps(rsq12,rcutoff2);
 985
 986             /* Update potential sum for this i atom from the interaction with this j atom. */
 987             velec            = _mm_and_ps(velec,cutoff_mask);
 988             velec            = _mm_andnot_ps(dummy_mask,velec);
 989             velecsum         = _mm_add_ps(velecsum,velec);
 990
 991             fscal            = felec;
 992
 993             fscal            = _mm_and_ps(fscal,cutoff_mask);
 994
 995             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 996
 997              /* Update vectorial force */
 998             fix1             = _mm_macc_ps(dx12,fscal,fix1);
 999             fiy1             = _mm_macc_ps(dy12,fscal,fiy1);
1000             fiz1             = _mm_macc_ps(dz12,fscal,fiz1);
1001
1002             fjx2             = _mm_macc_ps(dx12,fscal,fjx2);
1003             fjy2             = _mm_macc_ps(dy12,fscal,fjy2);
1004             fjz2             = _mm_macc_ps(dz12,fscal,fjz2);
1005
1006             }
1007
1008             /**************************
1009              * CALCULATE INTERACTIONS *
1010              **************************/
1011
1012             if (gmx_mm_any_lt(rsq13,rcutoff2))
1013             {
1014
1015             r13              = _mm_mul_ps(rsq13,rinv13);
1016             r13              = _mm_andnot_ps(dummy_mask,r13);
1017
1018             /* EWALD ELECTROSTATICS */
1019
1020             /* Analytical PME correction */
1021             zeta2            = _mm_mul_ps(beta2,rsq13);
1022             rinv3            = _mm_mul_ps(rinvsq13,rinv13);
1023             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1024             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1025             felec            = _mm_mul_ps(qq13,felec);
1026             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1027             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv13);
1028             velec            = _mm_mul_ps(qq13,velec);
1029
1030             d                = _mm_sub_ps(r13,rswitch);
1031             d                = _mm_max_ps(d,_mm_setzero_ps());
1032             d2               = _mm_mul_ps(d,d);
1033             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1034
1035             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1036
1037             /* Evaluate switch function */
1038             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1039             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv13,_mm_mul_ps(velec,dsw)) );
1040             velec            = _mm_mul_ps(velec,sw);
1041             cutoff_mask      = _mm_cmplt_ps(rsq13,rcutoff2);
1042
1043             /* Update potential sum for this i atom from the interaction with this j atom. */
1044             velec            = _mm_and_ps(velec,cutoff_mask);
1045             velec            = _mm_andnot_ps(dummy_mask,velec);
1046             velecsum         = _mm_add_ps(velecsum,velec);
1047
1048             fscal            = felec;
1049
1050             fscal            = _mm_and_ps(fscal,cutoff_mask);
1051
1052             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1053
1054              /* Update vectorial force */
1055             fix1             = _mm_macc_ps(dx13,fscal,fix1);
1056             fiy1             = _mm_macc_ps(dy13,fscal,fiy1);
1057             fiz1             = _mm_macc_ps(dz13,fscal,fiz1);
1058
1059             fjx3             = _mm_macc_ps(dx13,fscal,fjx3);
1060             fjy3             = _mm_macc_ps(dy13,fscal,fjy3);
1061             fjz3             = _mm_macc_ps(dz13,fscal,fjz3);
1062
1063             }
1064
1065             /**************************
1066              * CALCULATE INTERACTIONS *
1067              **************************/
1068
1069             if (gmx_mm_any_lt(rsq21,rcutoff2))
1070             {
1071
1072             r21              = _mm_mul_ps(rsq21,rinv21);
1073             r21              = _mm_andnot_ps(dummy_mask,r21);
1074
1075             /* EWALD ELECTROSTATICS */
1076
1077             /* Analytical PME correction */
1078             zeta2            = _mm_mul_ps(beta2,rsq21);
1079             rinv3            = _mm_mul_ps(rinvsq21,rinv21);
1080             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1081             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1082             felec            = _mm_mul_ps(qq21,felec);
1083             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1084             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv21);
1085             velec            = _mm_mul_ps(qq21,velec);
1086
1087             d                = _mm_sub_ps(r21,rswitch);
1088             d                = _mm_max_ps(d,_mm_setzero_ps());
1089             d2               = _mm_mul_ps(d,d);
1090             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1091
1092             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1093
1094             /* Evaluate switch function */
1095             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1096             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv21,_mm_mul_ps(velec,dsw)) );
1097             velec            = _mm_mul_ps(velec,sw);
1098             cutoff_mask      = _mm_cmplt_ps(rsq21,rcutoff2);
1099
1100             /* Update potential sum for this i atom from the interaction with this j atom. */
1101             velec            = _mm_and_ps(velec,cutoff_mask);
1102             velec            = _mm_andnot_ps(dummy_mask,velec);
1103             velecsum         = _mm_add_ps(velecsum,velec);
1104
1105             fscal            = felec;
1106
1107             fscal            = _mm_and_ps(fscal,cutoff_mask);
1108
1109             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1110
1111              /* Update vectorial force */
1112             fix2             = _mm_macc_ps(dx21,fscal,fix2);
1113             fiy2             = _mm_macc_ps(dy21,fscal,fiy2);
1114             fiz2             = _mm_macc_ps(dz21,fscal,fiz2);
1115
1116             fjx1             = _mm_macc_ps(dx21,fscal,fjx1);
1117             fjy1             = _mm_macc_ps(dy21,fscal,fjy1);
1118             fjz1             = _mm_macc_ps(dz21,fscal,fjz1);
1119
1120             }
1121
1122             /**************************
1123              * CALCULATE INTERACTIONS *
1124              **************************/
1125
1126             if (gmx_mm_any_lt(rsq22,rcutoff2))
1127             {
1128
1129             r22              = _mm_mul_ps(rsq22,rinv22);
1130             r22              = _mm_andnot_ps(dummy_mask,r22);
1131
1132             /* EWALD ELECTROSTATICS */
1133
1134             /* Analytical PME correction */
1135             zeta2            = _mm_mul_ps(beta2,rsq22);
1136             rinv3            = _mm_mul_ps(rinvsq22,rinv22);
1137             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1138             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1139             felec            = _mm_mul_ps(qq22,felec);
1140             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1141             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv22);
1142             velec            = _mm_mul_ps(qq22,velec);
1143
1144             d                = _mm_sub_ps(r22,rswitch);
1145             d                = _mm_max_ps(d,_mm_setzero_ps());
1146             d2               = _mm_mul_ps(d,d);
1147             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1148
1149             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1150
1151             /* Evaluate switch function */
1152             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1153             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv22,_mm_mul_ps(velec,dsw)) );
1154             velec            = _mm_mul_ps(velec,sw);
1155             cutoff_mask      = _mm_cmplt_ps(rsq22,rcutoff2);
1156
1157             /* Update potential sum for this i atom from the interaction with this j atom. */
1158             velec            = _mm_and_ps(velec,cutoff_mask);
1159             velec            = _mm_andnot_ps(dummy_mask,velec);
1160             velecsum         = _mm_add_ps(velecsum,velec);
1161
1162             fscal            = felec;
1163
1164             fscal            = _mm_and_ps(fscal,cutoff_mask);
1165
1166             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1167
1168              /* Update vectorial force */
1169             fix2             = _mm_macc_ps(dx22,fscal,fix2);
1170             fiy2             = _mm_macc_ps(dy22,fscal,fiy2);
1171             fiz2             = _mm_macc_ps(dz22,fscal,fiz2);
1172
1173             fjx2             = _mm_macc_ps(dx22,fscal,fjx2);
1174             fjy2             = _mm_macc_ps(dy22,fscal,fjy2);
1175             fjz2             = _mm_macc_ps(dz22,fscal,fjz2);
1176
1177             }
1178
1179             /**************************
1180              * CALCULATE INTERACTIONS *
1181              **************************/
1182
1183             if (gmx_mm_any_lt(rsq23,rcutoff2))
1184             {
1185
1186             r23              = _mm_mul_ps(rsq23,rinv23);
1187             r23              = _mm_andnot_ps(dummy_mask,r23);
1188
1189             /* EWALD ELECTROSTATICS */
1190
1191             /* Analytical PME correction */
1192             zeta2            = _mm_mul_ps(beta2,rsq23);
1193             rinv3            = _mm_mul_ps(rinvsq23,rinv23);
1194             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1195             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1196             felec            = _mm_mul_ps(qq23,felec);
1197             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1198             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv23);
1199             velec            = _mm_mul_ps(qq23,velec);
1200
1201             d                = _mm_sub_ps(r23,rswitch);
1202             d                = _mm_max_ps(d,_mm_setzero_ps());
1203             d2               = _mm_mul_ps(d,d);
1204             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1205
1206             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1207
1208             /* Evaluate switch function */
1209             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1210             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv23,_mm_mul_ps(velec,dsw)) );
1211             velec            = _mm_mul_ps(velec,sw);
1212             cutoff_mask      = _mm_cmplt_ps(rsq23,rcutoff2);
1213
1214             /* Update potential sum for this i atom from the interaction with this j atom. */
1215             velec            = _mm_and_ps(velec,cutoff_mask);
1216             velec            = _mm_andnot_ps(dummy_mask,velec);
1217             velecsum         = _mm_add_ps(velecsum,velec);
1218
1219             fscal            = felec;
1220
1221             fscal            = _mm_and_ps(fscal,cutoff_mask);
1222
1223             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1224
1225              /* Update vectorial force */
1226             fix2             = _mm_macc_ps(dx23,fscal,fix2);
1227             fiy2             = _mm_macc_ps(dy23,fscal,fiy2);
1228             fiz2             = _mm_macc_ps(dz23,fscal,fiz2);
1229
1230             fjx3             = _mm_macc_ps(dx23,fscal,fjx3);
1231             fjy3             = _mm_macc_ps(dy23,fscal,fjy3);
1232             fjz3             = _mm_macc_ps(dz23,fscal,fjz3);
1233
1234             }
1235
1236             /**************************
1237              * CALCULATE INTERACTIONS *
1238              **************************/
1239
1240             if (gmx_mm_any_lt(rsq31,rcutoff2))
1241             {
1242
1243             r31              = _mm_mul_ps(rsq31,rinv31);
1244             r31              = _mm_andnot_ps(dummy_mask,r31);
1245
1246             /* EWALD ELECTROSTATICS */
1247
1248             /* Analytical PME correction */
1249             zeta2            = _mm_mul_ps(beta2,rsq31);
1250             rinv3            = _mm_mul_ps(rinvsq31,rinv31);
1251             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1252             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1253             felec            = _mm_mul_ps(qq31,felec);
1254             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1255             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv31);
1256             velec            = _mm_mul_ps(qq31,velec);
1257
1258             d                = _mm_sub_ps(r31,rswitch);
1259             d                = _mm_max_ps(d,_mm_setzero_ps());
1260             d2               = _mm_mul_ps(d,d);
1261             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1262
1263             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1264
1265             /* Evaluate switch function */
1266             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1267             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv31,_mm_mul_ps(velec,dsw)) );
1268             velec            = _mm_mul_ps(velec,sw);
1269             cutoff_mask      = _mm_cmplt_ps(rsq31,rcutoff2);
1270
1271             /* Update potential sum for this i atom from the interaction with this j atom. */
1272             velec            = _mm_and_ps(velec,cutoff_mask);
1273             velec            = _mm_andnot_ps(dummy_mask,velec);
1274             velecsum         = _mm_add_ps(velecsum,velec);
1275
1276             fscal            = felec;
1277
1278             fscal            = _mm_and_ps(fscal,cutoff_mask);
1279
1280             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1281
1282              /* Update vectorial force */
1283             fix3             = _mm_macc_ps(dx31,fscal,fix3);
1284             fiy3             = _mm_macc_ps(dy31,fscal,fiy3);
1285             fiz3             = _mm_macc_ps(dz31,fscal,fiz3);
1286
1287             fjx1             = _mm_macc_ps(dx31,fscal,fjx1);
1288             fjy1             = _mm_macc_ps(dy31,fscal,fjy1);
1289             fjz1             = _mm_macc_ps(dz31,fscal,fjz1);
1290
1291             }
1292
1293             /**************************
1294              * CALCULATE INTERACTIONS *
1295              **************************/
1296
1297             if (gmx_mm_any_lt(rsq32,rcutoff2))
1298             {
1299
1300             r32              = _mm_mul_ps(rsq32,rinv32);
1301             r32              = _mm_andnot_ps(dummy_mask,r32);
1302
1303             /* EWALD ELECTROSTATICS */
1304
1305             /* Analytical PME correction */
1306             zeta2            = _mm_mul_ps(beta2,rsq32);
1307             rinv3            = _mm_mul_ps(rinvsq32,rinv32);
1308             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1309             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1310             felec            = _mm_mul_ps(qq32,felec);
1311             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1312             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv32);
1313             velec            = _mm_mul_ps(qq32,velec);
1314
1315             d                = _mm_sub_ps(r32,rswitch);
1316             d                = _mm_max_ps(d,_mm_setzero_ps());
1317             d2               = _mm_mul_ps(d,d);
1318             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1319
1320             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1321
1322             /* Evaluate switch function */
1323             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1324             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv32,_mm_mul_ps(velec,dsw)) );
1325             velec            = _mm_mul_ps(velec,sw);
1326             cutoff_mask      = _mm_cmplt_ps(rsq32,rcutoff2);
1327
1328             /* Update potential sum for this i atom from the interaction with this j atom. */
1329             velec            = _mm_and_ps(velec,cutoff_mask);
1330             velec            = _mm_andnot_ps(dummy_mask,velec);
1331             velecsum         = _mm_add_ps(velecsum,velec);
1332
1333             fscal            = felec;
1334
1335             fscal            = _mm_and_ps(fscal,cutoff_mask);
1336
1337             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1338
1339              /* Update vectorial force */
1340             fix3             = _mm_macc_ps(dx32,fscal,fix3);
1341             fiy3             = _mm_macc_ps(dy32,fscal,fiy3);
1342             fiz3             = _mm_macc_ps(dz32,fscal,fiz3);
1343
1344             fjx2             = _mm_macc_ps(dx32,fscal,fjx2);
1345             fjy2             = _mm_macc_ps(dy32,fscal,fjy2);
1346             fjz2             = _mm_macc_ps(dz32,fscal,fjz2);
1347
1348             }
1349
1350             /**************************
1351              * CALCULATE INTERACTIONS *
1352              **************************/
1353
1354             if (gmx_mm_any_lt(rsq33,rcutoff2))
1355             {
1356
1357             r33              = _mm_mul_ps(rsq33,rinv33);
1358             r33              = _mm_andnot_ps(dummy_mask,r33);
1359
1360             /* EWALD ELECTROSTATICS */
1361
1362             /* Analytical PME correction */
1363             zeta2            = _mm_mul_ps(beta2,rsq33);
1364             rinv3            = _mm_mul_ps(rinvsq33,rinv33);
1365             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1366             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1367             felec            = _mm_mul_ps(qq33,felec);
1368             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1369             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv33);
1370             velec            = _mm_mul_ps(qq33,velec);
1371
1372             d                = _mm_sub_ps(r33,rswitch);
1373             d                = _mm_max_ps(d,_mm_setzero_ps());
1374             d2               = _mm_mul_ps(d,d);
1375             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1376
1377             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1378
1379             /* Evaluate switch function */
1380             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1381             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv33,_mm_mul_ps(velec,dsw)) );
1382             velec            = _mm_mul_ps(velec,sw);
1383             cutoff_mask      = _mm_cmplt_ps(rsq33,rcutoff2);
1384
1385             /* Update potential sum for this i atom from the interaction with this j atom. */
1386             velec            = _mm_and_ps(velec,cutoff_mask);
1387             velec            = _mm_andnot_ps(dummy_mask,velec);
1388             velecsum         = _mm_add_ps(velecsum,velec);
1389
1390             fscal            = felec;
1391
1392             fscal            = _mm_and_ps(fscal,cutoff_mask);
1393
1394             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1395
1396              /* Update vectorial force */
1397             fix3             = _mm_macc_ps(dx33,fscal,fix3);
1398             fiy3             = _mm_macc_ps(dy33,fscal,fiy3);
1399             fiz3             = _mm_macc_ps(dz33,fscal,fiz3);
1400
1401             fjx3             = _mm_macc_ps(dx33,fscal,fjx3);
1402             fjy3             = _mm_macc_ps(dy33,fscal,fjy3);
1403             fjz3             = _mm_macc_ps(dz33,fscal,fjz3);
1404
1405             }
1406
1407             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1408             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1409             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1410             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1411
1412             gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1413                                                    fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1414
1415             /* Inner loop uses 486 flops */
1416         }
1417
1418         /* End of innermost loop */
1419
1420         gmx_mm_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1421                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
1422
1423         ggid                        = gid[iidx];
1424         /* Update potential energies */
1425         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1426
1427         /* Increment number of inner iterations */
1428         inneriter                  += j_index_end - j_index_start;
1429
1430         /* Outer loop uses 19 flops */
1431     }
1432
1433     /* Increment number of outer iterations */
1434     outeriter        += nri;
1435
1436     /* Update outer/inner flops */
1437
1438     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_VF,outeriter*19 + inneriter*486);
1439 }
1440 /*
1441  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwNone_GeomW4W4_F_avx_128_fma_single
1442  * Electrostatics interaction: Ewald
1443  * VdW interaction:            None
1444  * Geometry:                   Water4-Water4
1445  * Calculate force/pot:        Force
1446  */
1447 void
1448 nb_kernel_ElecEwSw_VdwNone_GeomW4W4_F_avx_128_fma_single
1449                     (t_nblist                    * gmx_restrict       nlist,
1450                      rvec                        * gmx_restrict          xx,
1451                      rvec                        * gmx_restrict          ff,
1452                      t_forcerec                  * gmx_restrict          fr,
1453                      t_mdatoms                   * gmx_restrict     mdatoms,
1454                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1455                      t_nrnb                      * gmx_restrict        nrnb)
1456 {
1457     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1458      * just 0 for non-waters.
1459      * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
1460      * jnr indices corresponding to data put in the four positions in the SIMD register.
1461      */
1462     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1463     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1464     int              jnrA,jnrB,jnrC,jnrD;
1465     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1466     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1467     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1468     real             rcutoff_scalar;
1469     real             *shiftvec,*fshift,*x,*f;
1470     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1471     real             scratch[4*DIM];
1472     __m128           fscal,rcutoff,rcutoff2,jidxall;
1473     int              vdwioffset1;
1474     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1475     int              vdwioffset2;
1476     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1477     int              vdwioffset3;
1478     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1479     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1480     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1481     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1482     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1483     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1484     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1485     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1486     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1487     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1488     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1489     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1490     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1491     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1492     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1493     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1494     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
1495     real             *charge;
1496     __m128i          ewitab;
1497     __m128           ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1498     __m128           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1499     real             *ewtab;
1500     __m128           rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
1501     real             rswitch_scalar,d_scalar;
1502     __m128           dummy_mask,cutoff_mask;
1503     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1504     __m128           one     = _mm_set1_ps(1.0);
1505     __m128           two     = _mm_set1_ps(2.0);
1506     x                = xx[0];
1507     f                = ff[0];
1508
1509     nri              = nlist->nri;
1510     iinr             = nlist->iinr;
1511     jindex           = nlist->jindex;
1512     jjnr             = nlist->jjnr;
1513     shiftidx         = nlist->shift;
1514     gid              = nlist->gid;
1515     shiftvec         = fr->shift_vec[0];
1516     fshift           = fr->fshift[0];
1517     facel            = _mm_set1_ps(fr->epsfac);
1518     charge           = mdatoms->chargeA;
1519
1520     sh_ewald         = _mm_set1_ps(fr->ic->sh_ewald);
1521     beta             = _mm_set1_ps(fr->ic->ewaldcoeff_q);
1522     beta2            = _mm_mul_ps(beta,beta);
1523     beta3            = _mm_mul_ps(beta,beta2);
1524     ewtab            = fr->ic->tabq_coul_FDV0;
1525     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
1526     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
1527
1528     /* Setup water-specific parameters */
1529     inr              = nlist->iinr[0];
1530     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1531     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1532     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
1533
1534     jq1              = _mm_set1_ps(charge[inr+1]);
1535     jq2              = _mm_set1_ps(charge[inr+2]);
1536     jq3              = _mm_set1_ps(charge[inr+3]);
1537     qq11             = _mm_mul_ps(iq1,jq1);
1538     qq12             = _mm_mul_ps(iq1,jq2);
1539     qq13             = _mm_mul_ps(iq1,jq3);
1540     qq21             = _mm_mul_ps(iq2,jq1);
1541     qq22             = _mm_mul_ps(iq2,jq2);
1542     qq23             = _mm_mul_ps(iq2,jq3);
1543     qq31             = _mm_mul_ps(iq3,jq1);
1544     qq32             = _mm_mul_ps(iq3,jq2);
1545     qq33             = _mm_mul_ps(iq3,jq3);
1546
1547     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1548     rcutoff_scalar   = fr->rcoulomb;
1549     rcutoff          = _mm_set1_ps(rcutoff_scalar);
1550     rcutoff2         = _mm_mul_ps(rcutoff,rcutoff);
1551
1552     rswitch_scalar   = fr->rcoulomb_switch;
1553     rswitch          = _mm_set1_ps(rswitch_scalar);
1554     /* Setup switch parameters */
1555     d_scalar         = rcutoff_scalar-rswitch_scalar;
1556     d                = _mm_set1_ps(d_scalar);
1557     swV3             = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
1558     swV4             = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1559     swV5             = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1560     swF2             = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
1561     swF3             = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1562     swF4             = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1563
1564     /* Avoid stupid compiler warnings */
1565     jnrA = jnrB = jnrC = jnrD = 0;
1566     j_coord_offsetA = 0;
1567     j_coord_offsetB = 0;
1568     j_coord_offsetC = 0;
1569     j_coord_offsetD = 0;
1570
1571     outeriter        = 0;
1572     inneriter        = 0;
1573
1574     for(iidx=0;iidx<4*DIM;iidx++)
1575     {
1576         scratch[iidx] = 0.0;
1577     }
1578
1579     /* Start outer loop over neighborlists */
1580     for(iidx=0; iidx<nri; iidx++)
1581     {
1582         /* Load shift vector for this list */
1583         i_shift_offset   = DIM*shiftidx[iidx];
1584
1585         /* Load limits for loop over neighbors */
1586         j_index_start    = jindex[iidx];
1587         j_index_end      = jindex[iidx+1];
1588
1589         /* Get outer coordinate index */
1590         inr              = iinr[iidx];
1591         i_coord_offset   = DIM*inr;
1592
1593         /* Load i particle coords and add shift vector */
1594         gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
1595                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1596
1597         fix1             = _mm_setzero_ps();
1598         fiy1             = _mm_setzero_ps();
1599         fiz1             = _mm_setzero_ps();
1600         fix2             = _mm_setzero_ps();
1601         fiy2             = _mm_setzero_ps();
1602         fiz2             = _mm_setzero_ps();
1603         fix3             = _mm_setzero_ps();
1604         fiy3             = _mm_setzero_ps();
1605         fiz3             = _mm_setzero_ps();
1606
1607         /* Start inner kernel loop */
1608         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1609         {
1610
1611             /* Get j neighbor index, and coordinate index */
1612             jnrA             = jjnr[jidx];
1613             jnrB             = jjnr[jidx+1];
1614             jnrC             = jjnr[jidx+2];
1615             jnrD             = jjnr[jidx+3];
1616             j_coord_offsetA  = DIM*jnrA;
1617             j_coord_offsetB  = DIM*jnrB;
1618             j_coord_offsetC  = DIM*jnrC;
1619             j_coord_offsetD  = DIM*jnrD;
1620
1621             /* load j atom coordinates */
1622             gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1623                                               x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
1624                                               &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1625
1626             /* Calculate displacement vector */
1627             dx11             = _mm_sub_ps(ix1,jx1);
1628             dy11             = _mm_sub_ps(iy1,jy1);
1629             dz11             = _mm_sub_ps(iz1,jz1);
1630             dx12             = _mm_sub_ps(ix1,jx2);
1631             dy12             = _mm_sub_ps(iy1,jy2);
1632             dz12             = _mm_sub_ps(iz1,jz2);
1633             dx13             = _mm_sub_ps(ix1,jx3);
1634             dy13             = _mm_sub_ps(iy1,jy3);
1635             dz13             = _mm_sub_ps(iz1,jz3);
1636             dx21             = _mm_sub_ps(ix2,jx1);
1637             dy21             = _mm_sub_ps(iy2,jy1);
1638             dz21             = _mm_sub_ps(iz2,jz1);
1639             dx22             = _mm_sub_ps(ix2,jx2);
1640             dy22             = _mm_sub_ps(iy2,jy2);
1641             dz22             = _mm_sub_ps(iz2,jz2);
1642             dx23             = _mm_sub_ps(ix2,jx3);
1643             dy23             = _mm_sub_ps(iy2,jy3);
1644             dz23             = _mm_sub_ps(iz2,jz3);
1645             dx31             = _mm_sub_ps(ix3,jx1);
1646             dy31             = _mm_sub_ps(iy3,jy1);
1647             dz31             = _mm_sub_ps(iz3,jz1);
1648             dx32             = _mm_sub_ps(ix3,jx2);
1649             dy32             = _mm_sub_ps(iy3,jy2);
1650             dz32             = _mm_sub_ps(iz3,jz2);
1651             dx33             = _mm_sub_ps(ix3,jx3);
1652             dy33             = _mm_sub_ps(iy3,jy3);
1653             dz33             = _mm_sub_ps(iz3,jz3);
1654
1655             /* Calculate squared distance and things based on it */
1656             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1657             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1658             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1659             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1660             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1661             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1662             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1663             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1664             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1665
1666             rinv11           = gmx_mm_invsqrt_ps(rsq11);
1667             rinv12           = gmx_mm_invsqrt_ps(rsq12);
1668             rinv13           = gmx_mm_invsqrt_ps(rsq13);
1669             rinv21           = gmx_mm_invsqrt_ps(rsq21);
1670             rinv22           = gmx_mm_invsqrt_ps(rsq22);
1671             rinv23           = gmx_mm_invsqrt_ps(rsq23);
1672             rinv31           = gmx_mm_invsqrt_ps(rsq31);
1673             rinv32           = gmx_mm_invsqrt_ps(rsq32);
1674             rinv33           = gmx_mm_invsqrt_ps(rsq33);
1675
1676             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
1677             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
1678             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
1679             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
1680             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
1681             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
1682             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
1683             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
1684             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
1685
1686             fjx1             = _mm_setzero_ps();
1687             fjy1             = _mm_setzero_ps();
1688             fjz1             = _mm_setzero_ps();
1689             fjx2             = _mm_setzero_ps();
1690             fjy2             = _mm_setzero_ps();
1691             fjz2             = _mm_setzero_ps();
1692             fjx3             = _mm_setzero_ps();
1693             fjy3             = _mm_setzero_ps();
1694             fjz3             = _mm_setzero_ps();
1695
1696             /**************************
1697              * CALCULATE INTERACTIONS *
1698              **************************/
1699
1700             if (gmx_mm_any_lt(rsq11,rcutoff2))
1701             {
1702
1703             r11              = _mm_mul_ps(rsq11,rinv11);
1704
1705             /* EWALD ELECTROSTATICS */
1706
1707             /* Analytical PME correction */
1708             zeta2            = _mm_mul_ps(beta2,rsq11);
1709             rinv3            = _mm_mul_ps(rinvsq11,rinv11);
1710             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1711             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1712             felec            = _mm_mul_ps(qq11,felec);
1713             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1714             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv11);
1715             velec            = _mm_mul_ps(qq11,velec);
1716
1717             d                = _mm_sub_ps(r11,rswitch);
1718             d                = _mm_max_ps(d,_mm_setzero_ps());
1719             d2               = _mm_mul_ps(d,d);
1720             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1721
1722             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1723
1724             /* Evaluate switch function */
1725             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1726             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv11,_mm_mul_ps(velec,dsw)) );
1727             cutoff_mask      = _mm_cmplt_ps(rsq11,rcutoff2);
1728
1729             fscal            = felec;
1730
1731             fscal            = _mm_and_ps(fscal,cutoff_mask);
1732
1733              /* Update vectorial force */
1734             fix1             = _mm_macc_ps(dx11,fscal,fix1);
1735             fiy1             = _mm_macc_ps(dy11,fscal,fiy1);
1736             fiz1             = _mm_macc_ps(dz11,fscal,fiz1);
1737
1738             fjx1             = _mm_macc_ps(dx11,fscal,fjx1);
1739             fjy1             = _mm_macc_ps(dy11,fscal,fjy1);
1740             fjz1             = _mm_macc_ps(dz11,fscal,fjz1);
1741
1742             }
1743
1744             /**************************
1745              * CALCULATE INTERACTIONS *
1746              **************************/
1747
1748             if (gmx_mm_any_lt(rsq12,rcutoff2))
1749             {
1750
1751             r12              = _mm_mul_ps(rsq12,rinv12);
1752
1753             /* EWALD ELECTROSTATICS */
1754
1755             /* Analytical PME correction */
1756             zeta2            = _mm_mul_ps(beta2,rsq12);
1757             rinv3            = _mm_mul_ps(rinvsq12,rinv12);
1758             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1759             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1760             felec            = _mm_mul_ps(qq12,felec);
1761             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1762             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv12);
1763             velec            = _mm_mul_ps(qq12,velec);
1764
1765             d                = _mm_sub_ps(r12,rswitch);
1766             d                = _mm_max_ps(d,_mm_setzero_ps());
1767             d2               = _mm_mul_ps(d,d);
1768             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1769
1770             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1771
1772             /* Evaluate switch function */
1773             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1774             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv12,_mm_mul_ps(velec,dsw)) );
1775             cutoff_mask      = _mm_cmplt_ps(rsq12,rcutoff2);
1776
1777             fscal            = felec;
1778
1779             fscal            = _mm_and_ps(fscal,cutoff_mask);
1780
1781              /* Update vectorial force */
1782             fix1             = _mm_macc_ps(dx12,fscal,fix1);
1783             fiy1             = _mm_macc_ps(dy12,fscal,fiy1);
1784             fiz1             = _mm_macc_ps(dz12,fscal,fiz1);
1785
1786             fjx2             = _mm_macc_ps(dx12,fscal,fjx2);
1787             fjy2             = _mm_macc_ps(dy12,fscal,fjy2);
1788             fjz2             = _mm_macc_ps(dz12,fscal,fjz2);
1789
1790             }
1791
1792             /**************************
1793              * CALCULATE INTERACTIONS *
1794              **************************/
1795
1796             if (gmx_mm_any_lt(rsq13,rcutoff2))
1797             {
1798
1799             r13              = _mm_mul_ps(rsq13,rinv13);
1800
1801             /* EWALD ELECTROSTATICS */
1802
1803             /* Analytical PME correction */
1804             zeta2            = _mm_mul_ps(beta2,rsq13);
1805             rinv3            = _mm_mul_ps(rinvsq13,rinv13);
1806             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1807             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1808             felec            = _mm_mul_ps(qq13,felec);
1809             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1810             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv13);
1811             velec            = _mm_mul_ps(qq13,velec);
1812
1813             d                = _mm_sub_ps(r13,rswitch);
1814             d                = _mm_max_ps(d,_mm_setzero_ps());
1815             d2               = _mm_mul_ps(d,d);
1816             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1817
1818             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1819
1820             /* Evaluate switch function */
1821             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1822             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv13,_mm_mul_ps(velec,dsw)) );
1823             cutoff_mask      = _mm_cmplt_ps(rsq13,rcutoff2);
1824
1825             fscal            = felec;
1826
1827             fscal            = _mm_and_ps(fscal,cutoff_mask);
1828
1829              /* Update vectorial force */
1830             fix1             = _mm_macc_ps(dx13,fscal,fix1);
1831             fiy1             = _mm_macc_ps(dy13,fscal,fiy1);
1832             fiz1             = _mm_macc_ps(dz13,fscal,fiz1);
1833
1834             fjx3             = _mm_macc_ps(dx13,fscal,fjx3);
1835             fjy3             = _mm_macc_ps(dy13,fscal,fjy3);
1836             fjz3             = _mm_macc_ps(dz13,fscal,fjz3);
1837
1838             }
1839
1840             /**************************
1841              * CALCULATE INTERACTIONS *
1842              **************************/
1843
1844             if (gmx_mm_any_lt(rsq21,rcutoff2))
1845             {
1846
1847             r21              = _mm_mul_ps(rsq21,rinv21);
1848
1849             /* EWALD ELECTROSTATICS */
1850
1851             /* Analytical PME correction */
1852             zeta2            = _mm_mul_ps(beta2,rsq21);
1853             rinv3            = _mm_mul_ps(rinvsq21,rinv21);
1854             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1855             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1856             felec            = _mm_mul_ps(qq21,felec);
1857             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1858             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv21);
1859             velec            = _mm_mul_ps(qq21,velec);
1860
1861             d                = _mm_sub_ps(r21,rswitch);
1862             d                = _mm_max_ps(d,_mm_setzero_ps());
1863             d2               = _mm_mul_ps(d,d);
1864             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1865
1866             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1867
1868             /* Evaluate switch function */
1869             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1870             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv21,_mm_mul_ps(velec,dsw)) );
1871             cutoff_mask      = _mm_cmplt_ps(rsq21,rcutoff2);
1872
1873             fscal            = felec;
1874
1875             fscal            = _mm_and_ps(fscal,cutoff_mask);
1876
1877              /* Update vectorial force */
1878             fix2             = _mm_macc_ps(dx21,fscal,fix2);
1879             fiy2             = _mm_macc_ps(dy21,fscal,fiy2);
1880             fiz2             = _mm_macc_ps(dz21,fscal,fiz2);
1881
1882             fjx1             = _mm_macc_ps(dx21,fscal,fjx1);
1883             fjy1             = _mm_macc_ps(dy21,fscal,fjy1);
1884             fjz1             = _mm_macc_ps(dz21,fscal,fjz1);
1885
1886             }
1887
1888             /**************************
1889              * CALCULATE INTERACTIONS *
1890              **************************/
1891
1892             if (gmx_mm_any_lt(rsq22,rcutoff2))
1893             {
1894
1895             r22              = _mm_mul_ps(rsq22,rinv22);
1896
1897             /* EWALD ELECTROSTATICS */
1898
1899             /* Analytical PME correction */
1900             zeta2            = _mm_mul_ps(beta2,rsq22);
1901             rinv3            = _mm_mul_ps(rinvsq22,rinv22);
1902             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1903             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1904             felec            = _mm_mul_ps(qq22,felec);
1905             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1906             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv22);
1907             velec            = _mm_mul_ps(qq22,velec);
1908
1909             d                = _mm_sub_ps(r22,rswitch);
1910             d                = _mm_max_ps(d,_mm_setzero_ps());
1911             d2               = _mm_mul_ps(d,d);
1912             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1913
1914             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1915
1916             /* Evaluate switch function */
1917             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1918             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv22,_mm_mul_ps(velec,dsw)) );
1919             cutoff_mask      = _mm_cmplt_ps(rsq22,rcutoff2);
1920
1921             fscal            = felec;
1922
1923             fscal            = _mm_and_ps(fscal,cutoff_mask);
1924
1925              /* Update vectorial force */
1926             fix2             = _mm_macc_ps(dx22,fscal,fix2);
1927             fiy2             = _mm_macc_ps(dy22,fscal,fiy2);
1928             fiz2             = _mm_macc_ps(dz22,fscal,fiz2);
1929
1930             fjx2             = _mm_macc_ps(dx22,fscal,fjx2);
1931             fjy2             = _mm_macc_ps(dy22,fscal,fjy2);
1932             fjz2             = _mm_macc_ps(dz22,fscal,fjz2);
1933
1934             }
1935
1936             /**************************
1937              * CALCULATE INTERACTIONS *
1938              **************************/
1939
1940             if (gmx_mm_any_lt(rsq23,rcutoff2))
1941             {
1942
1943             r23              = _mm_mul_ps(rsq23,rinv23);
1944
1945             /* EWALD ELECTROSTATICS */
1946
1947             /* Analytical PME correction */
1948             zeta2            = _mm_mul_ps(beta2,rsq23);
1949             rinv3            = _mm_mul_ps(rinvsq23,rinv23);
1950             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1951             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1952             felec            = _mm_mul_ps(qq23,felec);
1953             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1954             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv23);
1955             velec            = _mm_mul_ps(qq23,velec);
1956
1957             d                = _mm_sub_ps(r23,rswitch);
1958             d                = _mm_max_ps(d,_mm_setzero_ps());
1959             d2               = _mm_mul_ps(d,d);
1960             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1961
1962             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1963
1964             /* Evaluate switch function */
1965             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1966             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv23,_mm_mul_ps(velec,dsw)) );
1967             cutoff_mask      = _mm_cmplt_ps(rsq23,rcutoff2);
1968
1969             fscal            = felec;
1970
1971             fscal            = _mm_and_ps(fscal,cutoff_mask);
1972
1973              /* Update vectorial force */
1974             fix2             = _mm_macc_ps(dx23,fscal,fix2);
1975             fiy2             = _mm_macc_ps(dy23,fscal,fiy2);
1976             fiz2             = _mm_macc_ps(dz23,fscal,fiz2);
1977
1978             fjx3             = _mm_macc_ps(dx23,fscal,fjx3);
1979             fjy3             = _mm_macc_ps(dy23,fscal,fjy3);
1980             fjz3             = _mm_macc_ps(dz23,fscal,fjz3);
1981
1982             }
1983
1984             /**************************
1985              * CALCULATE INTERACTIONS *
1986              **************************/
1987
1988             if (gmx_mm_any_lt(rsq31,rcutoff2))
1989             {
1990
1991             r31              = _mm_mul_ps(rsq31,rinv31);
1992
1993             /* EWALD ELECTROSTATICS */
1994
1995             /* Analytical PME correction */
1996             zeta2            = _mm_mul_ps(beta2,rsq31);
1997             rinv3            = _mm_mul_ps(rinvsq31,rinv31);
1998             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1999             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
2000             felec            = _mm_mul_ps(qq31,felec);
2001             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
2002             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv31);
2003             velec            = _mm_mul_ps(qq31,velec);
2004
2005             d                = _mm_sub_ps(r31,rswitch);
2006             d                = _mm_max_ps(d,_mm_setzero_ps());
2007             d2               = _mm_mul_ps(d,d);
2008             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
2009
2010             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
2011
2012             /* Evaluate switch function */
2013             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2014             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv31,_mm_mul_ps(velec,dsw)) );
2015             cutoff_mask      = _mm_cmplt_ps(rsq31,rcutoff2);
2016
2017             fscal            = felec;
2018
2019             fscal            = _mm_and_ps(fscal,cutoff_mask);
2020
2021              /* Update vectorial force */
2022             fix3             = _mm_macc_ps(dx31,fscal,fix3);
2023             fiy3             = _mm_macc_ps(dy31,fscal,fiy3);
2024             fiz3             = _mm_macc_ps(dz31,fscal,fiz3);
2025
2026             fjx1             = _mm_macc_ps(dx31,fscal,fjx1);
2027             fjy1             = _mm_macc_ps(dy31,fscal,fjy1);
2028             fjz1             = _mm_macc_ps(dz31,fscal,fjz1);
2029
2030             }
2031
2032             /**************************
2033              * CALCULATE INTERACTIONS *
2034              **************************/
2035
2036             if (gmx_mm_any_lt(rsq32,rcutoff2))
2037             {
2038
2039             r32              = _mm_mul_ps(rsq32,rinv32);
2040
2041             /* EWALD ELECTROSTATICS */
2042
2043             /* Analytical PME correction */
2044             zeta2            = _mm_mul_ps(beta2,rsq32);
2045             rinv3            = _mm_mul_ps(rinvsq32,rinv32);
2046             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
2047             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
2048             felec            = _mm_mul_ps(qq32,felec);
2049             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
2050             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv32);
2051             velec            = _mm_mul_ps(qq32,velec);
2052
2053             d                = _mm_sub_ps(r32,rswitch);
2054             d                = _mm_max_ps(d,_mm_setzero_ps());
2055             d2               = _mm_mul_ps(d,d);
2056             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
2057
2058             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
2059
2060             /* Evaluate switch function */
2061             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2062             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv32,_mm_mul_ps(velec,dsw)) );
2063             cutoff_mask      = _mm_cmplt_ps(rsq32,rcutoff2);
2064
2065             fscal            = felec;
2066
2067             fscal            = _mm_and_ps(fscal,cutoff_mask);
2068
2069              /* Update vectorial force */
2070             fix3             = _mm_macc_ps(dx32,fscal,fix3);
2071             fiy3             = _mm_macc_ps(dy32,fscal,fiy3);
2072             fiz3             = _mm_macc_ps(dz32,fscal,fiz3);
2073
2074             fjx2             = _mm_macc_ps(dx32,fscal,fjx2);
2075             fjy2             = _mm_macc_ps(dy32,fscal,fjy2);
2076             fjz2             = _mm_macc_ps(dz32,fscal,fjz2);
2077
2078             }
2079
2080             /**************************
2081              * CALCULATE INTERACTIONS *
2082              **************************/
2083
2084             if (gmx_mm_any_lt(rsq33,rcutoff2))
2085             {
2086
2087             r33              = _mm_mul_ps(rsq33,rinv33);
2088
2089             /* EWALD ELECTROSTATICS */
2090
2091             /* Analytical PME correction */
2092             zeta2            = _mm_mul_ps(beta2,rsq33);
2093             rinv3            = _mm_mul_ps(rinvsq33,rinv33);
2094             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
2095             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
2096             felec            = _mm_mul_ps(qq33,felec);
2097             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
2098             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv33);
2099             velec            = _mm_mul_ps(qq33,velec);
2100
2101             d                = _mm_sub_ps(r33,rswitch);
2102             d                = _mm_max_ps(d,_mm_setzero_ps());
2103             d2               = _mm_mul_ps(d,d);
2104             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
2105
2106             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
2107
2108             /* Evaluate switch function */
2109             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2110             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv33,_mm_mul_ps(velec,dsw)) );
2111             cutoff_mask      = _mm_cmplt_ps(rsq33,rcutoff2);
2112
2113             fscal            = felec;
2114
2115             fscal            = _mm_and_ps(fscal,cutoff_mask);
2116
2117              /* Update vectorial force */
2118             fix3             = _mm_macc_ps(dx33,fscal,fix3);
2119             fiy3             = _mm_macc_ps(dy33,fscal,fiy3);
2120             fiz3             = _mm_macc_ps(dz33,fscal,fiz3);
2121
2122             fjx3             = _mm_macc_ps(dx33,fscal,fjx3);
2123             fjy3             = _mm_macc_ps(dy33,fscal,fjy3);
2124             fjz3             = _mm_macc_ps(dz33,fscal,fjz3);
2125
2126             }
2127
2128             fjptrA             = f+j_coord_offsetA;
2129             fjptrB             = f+j_coord_offsetB;
2130             fjptrC             = f+j_coord_offsetC;
2131             fjptrD             = f+j_coord_offsetD;
2132
2133             gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
2134                                                    fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2135
2136             /* Inner loop uses 450 flops */
2137         }
2138
2139         if(jidx<j_index_end)
2140         {
2141
2142             /* Get j neighbor index, and coordinate index */
2143             jnrlistA         = jjnr[jidx];
2144             jnrlistB         = jjnr[jidx+1];
2145             jnrlistC         = jjnr[jidx+2];
2146             jnrlistD         = jjnr[jidx+3];
2147             /* Sign of each element will be negative for non-real atoms.
2148              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
2149              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
2150              */
2151             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
2152             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
2153             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
2154             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
2155             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
2156             j_coord_offsetA  = DIM*jnrA;
2157             j_coord_offsetB  = DIM*jnrB;
2158             j_coord_offsetC  = DIM*jnrC;
2159             j_coord_offsetD  = DIM*jnrD;
2160
2161             /* load j atom coordinates */
2162             gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
2163                                               x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
2164                                               &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
2165
2166             /* Calculate displacement vector */
2167             dx11             = _mm_sub_ps(ix1,jx1);
2168             dy11             = _mm_sub_ps(iy1,jy1);
2169             dz11             = _mm_sub_ps(iz1,jz1);
2170             dx12             = _mm_sub_ps(ix1,jx2);
2171             dy12             = _mm_sub_ps(iy1,jy2);
2172             dz12             = _mm_sub_ps(iz1,jz2);
2173             dx13             = _mm_sub_ps(ix1,jx3);
2174             dy13             = _mm_sub_ps(iy1,jy3);
2175             dz13             = _mm_sub_ps(iz1,jz3);
2176             dx21             = _mm_sub_ps(ix2,jx1);
2177             dy21             = _mm_sub_ps(iy2,jy1);
2178             dz21             = _mm_sub_ps(iz2,jz1);
2179             dx22             = _mm_sub_ps(ix2,jx2);
2180             dy22             = _mm_sub_ps(iy2,jy2);
2181             dz22             = _mm_sub_ps(iz2,jz2);
2182             dx23             = _mm_sub_ps(ix2,jx3);
2183             dy23             = _mm_sub_ps(iy2,jy3);
2184             dz23             = _mm_sub_ps(iz2,jz3);
2185             dx31             = _mm_sub_ps(ix3,jx1);
2186             dy31             = _mm_sub_ps(iy3,jy1);
2187             dz31             = _mm_sub_ps(iz3,jz1);
2188             dx32             = _mm_sub_ps(ix3,jx2);
2189             dy32             = _mm_sub_ps(iy3,jy2);
2190             dz32             = _mm_sub_ps(iz3,jz2);
2191             dx33             = _mm_sub_ps(ix3,jx3);
2192             dy33             = _mm_sub_ps(iy3,jy3);
2193             dz33             = _mm_sub_ps(iz3,jz3);
2194
2195             /* Calculate squared distance and things based on it */
2196             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
2197             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
2198             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
2199             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
2200             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
2201             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
2202             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
2203             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
2204             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
2205
2206             rinv11           = gmx_mm_invsqrt_ps(rsq11);
2207             rinv12           = gmx_mm_invsqrt_ps(rsq12);
2208             rinv13           = gmx_mm_invsqrt_ps(rsq13);
2209             rinv21           = gmx_mm_invsqrt_ps(rsq21);
2210             rinv22           = gmx_mm_invsqrt_ps(rsq22);
2211             rinv23           = gmx_mm_invsqrt_ps(rsq23);
2212             rinv31           = gmx_mm_invsqrt_ps(rsq31);
2213             rinv32           = gmx_mm_invsqrt_ps(rsq32);
2214             rinv33           = gmx_mm_invsqrt_ps(rsq33);
2215
2216             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
2217             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
2218             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
2219             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
2220             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
2221             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
2222             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
2223             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
2224             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
2225
2226             fjx1             = _mm_setzero_ps();
2227             fjy1             = _mm_setzero_ps();
2228             fjz1             = _mm_setzero_ps();
2229             fjx2             = _mm_setzero_ps();
2230             fjy2             = _mm_setzero_ps();
2231             fjz2             = _mm_setzero_ps();
2232             fjx3             = _mm_setzero_ps();
2233             fjy3             = _mm_setzero_ps();
2234             fjz3             = _mm_setzero_ps();
2235
2236             /**************************
2237              * CALCULATE INTERACTIONS *
2238              **************************/
2239
2240             if (gmx_mm_any_lt(rsq11,rcutoff2))
2241             {
2242
2243             r11              = _mm_mul_ps(rsq11,rinv11);
2244             r11              = _mm_andnot_ps(dummy_mask,r11);
2245
2246             /* EWALD ELECTROSTATICS */
2247
2248             /* Analytical PME correction */
2249             zeta2            = _mm_mul_ps(beta2,rsq11);
2250             rinv3            = _mm_mul_ps(rinvsq11,rinv11);
2251             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
2252             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
2253             felec            = _mm_mul_ps(qq11,felec);
2254             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
2255             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv11);
2256             velec            = _mm_mul_ps(qq11,velec);
2257
2258             d                = _mm_sub_ps(r11,rswitch);
2259             d                = _mm_max_ps(d,_mm_setzero_ps());
2260             d2               = _mm_mul_ps(d,d);
2261             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
2262
2263             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
2264
2265             /* Evaluate switch function */
2266             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2267             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv11,_mm_mul_ps(velec,dsw)) );
2268             cutoff_mask      = _mm_cmplt_ps(rsq11,rcutoff2);
2269
2270             fscal            = felec;
2271
2272             fscal            = _mm_and_ps(fscal,cutoff_mask);
2273
2274             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2275
2276              /* Update vectorial force */
2277             fix1             = _mm_macc_ps(dx11,fscal,fix1);
2278             fiy1             = _mm_macc_ps(dy11,fscal,fiy1);
2279             fiz1             = _mm_macc_ps(dz11,fscal,fiz1);
2280
2281             fjx1             = _mm_macc_ps(dx11,fscal,fjx1);
2282             fjy1             = _mm_macc_ps(dy11,fscal,fjy1);
2283             fjz1             = _mm_macc_ps(dz11,fscal,fjz1);
2284
2285             }
2286
2287             /**************************
2288              * CALCULATE INTERACTIONS *
2289              **************************/
2290
2291             if (gmx_mm_any_lt(rsq12,rcutoff2))
2292             {
2293
2294             r12              = _mm_mul_ps(rsq12,rinv12);
2295             r12              = _mm_andnot_ps(dummy_mask,r12);
2296
2297             /* EWALD ELECTROSTATICS */
2298
2299             /* Analytical PME correction */
2300             zeta2            = _mm_mul_ps(beta2,rsq12);
2301             rinv3            = _mm_mul_ps(rinvsq12,rinv12);
2302             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
2303             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
2304             felec            = _mm_mul_ps(qq12,felec);
2305             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
2306             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv12);
2307             velec            = _mm_mul_ps(qq12,velec);
2308
2309             d                = _mm_sub_ps(r12,rswitch);
2310             d                = _mm_max_ps(d,_mm_setzero_ps());
2311             d2               = _mm_mul_ps(d,d);
2312             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
2313
2314             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
2315
2316             /* Evaluate switch function */
2317             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2318             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv12,_mm_mul_ps(velec,dsw)) );
2319             cutoff_mask      = _mm_cmplt_ps(rsq12,rcutoff2);
2320
2321             fscal            = felec;
2322
2323             fscal            = _mm_and_ps(fscal,cutoff_mask);
2324
2325             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2326
2327              /* Update vectorial force */
2328             fix1             = _mm_macc_ps(dx12,fscal,fix1);
2329             fiy1             = _mm_macc_ps(dy12,fscal,fiy1);
2330             fiz1             = _mm_macc_ps(dz12,fscal,fiz1);
2331
2332             fjx2             = _mm_macc_ps(dx12,fscal,fjx2);
2333             fjy2             = _mm_macc_ps(dy12,fscal,fjy2);
2334             fjz2             = _mm_macc_ps(dz12,fscal,fjz2);
2335
2336             }
2337
2338             /**************************
2339              * CALCULATE INTERACTIONS *
2340              **************************/
2341
2342             if (gmx_mm_any_lt(rsq13,rcutoff2))
2343             {
2344
2345             r13              = _mm_mul_ps(rsq13,rinv13);
2346             r13              = _mm_andnot_ps(dummy_mask,r13);
2347
2348             /* EWALD ELECTROSTATICS */
2349
2350             /* Analytical PME correction */
2351             zeta2            = _mm_mul_ps(beta2,rsq13);
2352             rinv3            = _mm_mul_ps(rinvsq13,rinv13);
2353             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
2354             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
2355             felec            = _mm_mul_ps(qq13,felec);
2356             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
2357             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv13);
2358             velec            = _mm_mul_ps(qq13,velec);
2359
2360             d                = _mm_sub_ps(r13,rswitch);
2361             d                = _mm_max_ps(d,_mm_setzero_ps());
2362             d2               = _mm_mul_ps(d,d);
2363             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
2364
2365             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
2366
2367             /* Evaluate switch function */
2368             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2369             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv13,_mm_mul_ps(velec,dsw)) );
2370             cutoff_mask      = _mm_cmplt_ps(rsq13,rcutoff2);
2371
2372             fscal            = felec;
2373
2374             fscal            = _mm_and_ps(fscal,cutoff_mask);
2375
2376             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2377
2378              /* Update vectorial force */
2379             fix1             = _mm_macc_ps(dx13,fscal,fix1);
2380             fiy1             = _mm_macc_ps(dy13,fscal,fiy1);
2381             fiz1             = _mm_macc_ps(dz13,fscal,fiz1);
2382
2383             fjx3             = _mm_macc_ps(dx13,fscal,fjx3);
2384             fjy3             = _mm_macc_ps(dy13,fscal,fjy3);
2385             fjz3             = _mm_macc_ps(dz13,fscal,fjz3);
2386
2387             }
2388
2389             /**************************
2390              * CALCULATE INTERACTIONS *
2391              **************************/
2392
2393             if (gmx_mm_any_lt(rsq21,rcutoff2))
2394             {
2395
2396             r21              = _mm_mul_ps(rsq21,rinv21);
2397             r21              = _mm_andnot_ps(dummy_mask,r21);
2398
2399             /* EWALD ELECTROSTATICS */
2400
2401             /* Analytical PME correction */
2402             zeta2            = _mm_mul_ps(beta2,rsq21);
2403             rinv3            = _mm_mul_ps(rinvsq21,rinv21);
2404             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
2405             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
2406             felec            = _mm_mul_ps(qq21,felec);
2407             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
2408             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv21);
2409             velec            = _mm_mul_ps(qq21,velec);
2410
2411             d                = _mm_sub_ps(r21,rswitch);
2412             d                = _mm_max_ps(d,_mm_setzero_ps());
2413             d2               = _mm_mul_ps(d,d);
2414             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
2415
2416             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
2417
2418             /* Evaluate switch function */
2419             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2420             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv21,_mm_mul_ps(velec,dsw)) );
2421             cutoff_mask      = _mm_cmplt_ps(rsq21,rcutoff2);
2422
2423             fscal            = felec;
2424
2425             fscal            = _mm_and_ps(fscal,cutoff_mask);
2426
2427             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2428
2429              /* Update vectorial force */
2430             fix2             = _mm_macc_ps(dx21,fscal,fix2);
2431             fiy2             = _mm_macc_ps(dy21,fscal,fiy2);
2432             fiz2             = _mm_macc_ps(dz21,fscal,fiz2);
2433
2434             fjx1             = _mm_macc_ps(dx21,fscal,fjx1);
2435             fjy1             = _mm_macc_ps(dy21,fscal,fjy1);
2436             fjz1             = _mm_macc_ps(dz21,fscal,fjz1);
2437
2438             }
2439
2440             /**************************
2441              * CALCULATE INTERACTIONS *
2442              **************************/
2443
2444             if (gmx_mm_any_lt(rsq22,rcutoff2))
2445             {
2446
2447             r22              = _mm_mul_ps(rsq22,rinv22);
2448             r22              = _mm_andnot_ps(dummy_mask,r22);
2449
2450             /* EWALD ELECTROSTATICS */
2451
2452             /* Analytical PME correction */
2453             zeta2            = _mm_mul_ps(beta2,rsq22);
2454             rinv3            = _mm_mul_ps(rinvsq22,rinv22);
2455             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
2456             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
2457             felec            = _mm_mul_ps(qq22,felec);
2458             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
2459             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv22);
2460             velec            = _mm_mul_ps(qq22,velec);
2461
2462             d                = _mm_sub_ps(r22,rswitch);
2463             d                = _mm_max_ps(d,_mm_setzero_ps());
2464             d2               = _mm_mul_ps(d,d);
2465             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
2466
2467             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
2468
2469             /* Evaluate switch function */
2470             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2471             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv22,_mm_mul_ps(velec,dsw)) );
2472             cutoff_mask      = _mm_cmplt_ps(rsq22,rcutoff2);
2473
2474             fscal            = felec;
2475
2476             fscal            = _mm_and_ps(fscal,cutoff_mask);
2477
2478             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2479
2480              /* Update vectorial force */
2481             fix2             = _mm_macc_ps(dx22,fscal,fix2);
2482             fiy2             = _mm_macc_ps(dy22,fscal,fiy2);
2483             fiz2             = _mm_macc_ps(dz22,fscal,fiz2);
2484
2485             fjx2             = _mm_macc_ps(dx22,fscal,fjx2);
2486             fjy2             = _mm_macc_ps(dy22,fscal,fjy2);
2487             fjz2             = _mm_macc_ps(dz22,fscal,fjz2);
2488
2489             }
2490
2491             /**************************
2492              * CALCULATE INTERACTIONS *
2493              **************************/
2494
2495             if (gmx_mm_any_lt(rsq23,rcutoff2))
2496             {
2497
2498             r23              = _mm_mul_ps(rsq23,rinv23);
2499             r23              = _mm_andnot_ps(dummy_mask,r23);
2500
2501             /* EWALD ELECTROSTATICS */
2502
2503             /* Analytical PME correction */
2504             zeta2            = _mm_mul_ps(beta2,rsq23);
2505             rinv3            = _mm_mul_ps(rinvsq23,rinv23);
2506             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
2507             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
2508             felec            = _mm_mul_ps(qq23,felec);
2509             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
2510             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv23);
2511             velec            = _mm_mul_ps(qq23,velec);
2512
2513             d                = _mm_sub_ps(r23,rswitch);
2514             d                = _mm_max_ps(d,_mm_setzero_ps());
2515             d2               = _mm_mul_ps(d,d);
2516             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
2517
2518             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
2519
2520             /* Evaluate switch function */
2521             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2522             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv23,_mm_mul_ps(velec,dsw)) );
2523             cutoff_mask      = _mm_cmplt_ps(rsq23,rcutoff2);
2524
2525             fscal            = felec;
2526
2527             fscal            = _mm_and_ps(fscal,cutoff_mask);
2528
2529             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2530
2531              /* Update vectorial force */
2532             fix2             = _mm_macc_ps(dx23,fscal,fix2);
2533             fiy2             = _mm_macc_ps(dy23,fscal,fiy2);
2534             fiz2             = _mm_macc_ps(dz23,fscal,fiz2);
2535
2536             fjx3             = _mm_macc_ps(dx23,fscal,fjx3);
2537             fjy3             = _mm_macc_ps(dy23,fscal,fjy3);
2538             fjz3             = _mm_macc_ps(dz23,fscal,fjz3);
2539
2540             }
2541
2542             /**************************
2543              * CALCULATE INTERACTIONS *
2544              **************************/
2545
2546             if (gmx_mm_any_lt(rsq31,rcutoff2))
2547             {
2548
2549             r31              = _mm_mul_ps(rsq31,rinv31);
2550             r31              = _mm_andnot_ps(dummy_mask,r31);
2551
2552             /* EWALD ELECTROSTATICS */
2553
2554             /* Analytical PME correction */
2555             zeta2            = _mm_mul_ps(beta2,rsq31);
2556             rinv3            = _mm_mul_ps(rinvsq31,rinv31);
2557             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
2558             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
2559             felec            = _mm_mul_ps(qq31,felec);
2560             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
2561             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv31);
2562             velec            = _mm_mul_ps(qq31,velec);
2563
2564             d                = _mm_sub_ps(r31,rswitch);
2565             d                = _mm_max_ps(d,_mm_setzero_ps());
2566             d2               = _mm_mul_ps(d,d);
2567             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
2568
2569             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
2570
2571             /* Evaluate switch function */
2572             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2573             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv31,_mm_mul_ps(velec,dsw)) );
2574             cutoff_mask      = _mm_cmplt_ps(rsq31,rcutoff2);
2575
2576             fscal            = felec;
2577
2578             fscal            = _mm_and_ps(fscal,cutoff_mask);
2579
2580             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2581
2582              /* Update vectorial force */
2583             fix3             = _mm_macc_ps(dx31,fscal,fix3);
2584             fiy3             = _mm_macc_ps(dy31,fscal,fiy3);
2585             fiz3             = _mm_macc_ps(dz31,fscal,fiz3);
2586
2587             fjx1             = _mm_macc_ps(dx31,fscal,fjx1);
2588             fjy1             = _mm_macc_ps(dy31,fscal,fjy1);
2589             fjz1             = _mm_macc_ps(dz31,fscal,fjz1);
2590
2591             }
2592
2593             /**************************
2594              * CALCULATE INTERACTIONS *
2595              **************************/
2596
2597             if (gmx_mm_any_lt(rsq32,rcutoff2))
2598             {
2599
2600             r32              = _mm_mul_ps(rsq32,rinv32);
2601             r32              = _mm_andnot_ps(dummy_mask,r32);
2602
2603             /* EWALD ELECTROSTATICS */
2604
2605             /* Analytical PME correction */
2606             zeta2            = _mm_mul_ps(beta2,rsq32);
2607             rinv3            = _mm_mul_ps(rinvsq32,rinv32);
2608             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
2609             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
2610             felec            = _mm_mul_ps(qq32,felec);
2611             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
2612             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv32);
2613             velec            = _mm_mul_ps(qq32,velec);
2614
2615             d                = _mm_sub_ps(r32,rswitch);
2616             d                = _mm_max_ps(d,_mm_setzero_ps());
2617             d2               = _mm_mul_ps(d,d);
2618             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
2619
2620             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
2621
2622             /* Evaluate switch function */
2623             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2624             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv32,_mm_mul_ps(velec,dsw)) );
2625             cutoff_mask      = _mm_cmplt_ps(rsq32,rcutoff2);
2626
2627             fscal            = felec;
2628
2629             fscal            = _mm_and_ps(fscal,cutoff_mask);
2630
2631             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2632
2633              /* Update vectorial force */
2634             fix3             = _mm_macc_ps(dx32,fscal,fix3);
2635             fiy3             = _mm_macc_ps(dy32,fscal,fiy3);
2636             fiz3             = _mm_macc_ps(dz32,fscal,fiz3);
2637
2638             fjx2             = _mm_macc_ps(dx32,fscal,fjx2);
2639             fjy2             = _mm_macc_ps(dy32,fscal,fjy2);
2640             fjz2             = _mm_macc_ps(dz32,fscal,fjz2);
2641
2642             }
2643
2644             /**************************
2645              * CALCULATE INTERACTIONS *
2646              **************************/
2647
2648             if (gmx_mm_any_lt(rsq33,rcutoff2))
2649             {
2650
2651             r33              = _mm_mul_ps(rsq33,rinv33);
2652             r33              = _mm_andnot_ps(dummy_mask,r33);
2653
2654             /* EWALD ELECTROSTATICS */
2655
2656             /* Analytical PME correction */
2657             zeta2            = _mm_mul_ps(beta2,rsq33);
2658             rinv3            = _mm_mul_ps(rinvsq33,rinv33);
2659             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
2660             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
2661             felec            = _mm_mul_ps(qq33,felec);
2662             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
2663             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv33);
2664             velec            = _mm_mul_ps(qq33,velec);
2665
2666             d                = _mm_sub_ps(r33,rswitch);
2667             d                = _mm_max_ps(d,_mm_setzero_ps());
2668             d2               = _mm_mul_ps(d,d);
2669             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
2670
2671             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
2672
2673             /* Evaluate switch function */
2674             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2675             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv33,_mm_mul_ps(velec,dsw)) );
2676             cutoff_mask      = _mm_cmplt_ps(rsq33,rcutoff2);
2677
2678             fscal            = felec;
2679
2680             fscal            = _mm_and_ps(fscal,cutoff_mask);
2681
2682             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2683
2684              /* Update vectorial force */
2685             fix3             = _mm_macc_ps(dx33,fscal,fix3);
2686             fiy3             = _mm_macc_ps(dy33,fscal,fiy3);
2687             fiz3             = _mm_macc_ps(dz33,fscal,fiz3);
2688
2689             fjx3             = _mm_macc_ps(dx33,fscal,fjx3);
2690             fjy3             = _mm_macc_ps(dy33,fscal,fjy3);
2691             fjz3             = _mm_macc_ps(dz33,fscal,fjz3);
2692
2693             }
2694
2695             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2696             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2697             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2698             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2699
2700             gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
2701                                                    fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2702
2703             /* Inner loop uses 459 flops */
2704         }
2705
2706         /* End of innermost loop */
2707
2708         gmx_mm_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2709                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
2710
2711         /* Increment number of inner iterations */
2712         inneriter                  += j_index_end - j_index_start;
2713
2714         /* Outer loop uses 18 flops */
2715     }
2716
2717     /* Increment number of outer iterations */
2718     outeriter        += nri;
2719
2720     /* Update outer/inner flops */
2721
2722     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_F,outeriter*18 + inneriter*459);
2723 }